Recommendations for the analysis of the effect of treatment on the development of second malignancies

Use of the person-years method for evaluating the association between treatment of a primary cancer and subsequent development of a second malignancy is reviewed. For this type of analysis, the risk of developing a second malignancy is implicitly assumed to remain constant during each patient's follow-up period; this assumption is shown to be inappropriate. When the oncogenic potential of two or more treatments is compared, results are biased unfavorably against the use of intensive treatments that prolong life. Misinterpretation of the oncogenic potential of such treatment regimens can therefore occur, and two alternative statistical techniques are proposed, each for use in a commonly encountered experimental situation. These fairly classic methods of survival analysis are recommended to ascertain the relationship between treatment and development of a second cancer.